Neural network for forecasting exchange rates

Education work

Configuration file

1. Layers count N
2. N numbers - neuron count on each layer
3. If next data it's neuron weight then 1 else 0
4. i line: synapse count on current neuron M
5. i+1 line: M numbers - each neuron weight

Compile and launch

Unpack archive and run this commands

$ cd neural-network/core/  
$ make prepare  
$ make  

For launch use

$ ./bin/nn  

Parameters list

Long	Short	Meaning	Required params
--version	-v	Print version and developer	None
--help	-h	Print help	None
--learn	-l	Launch neural network for learning	-s -d -i -o
--test	-t	Testing neural network	-d -i
--process	-p	Processing neural network	-d -i
--input file	-i file	Input file with scheme (and weights for neurons)	None
--output file	-o file	output file for learning, testing or processing result	None
--data file	-d file	Data file for learning, testing or processing neural network	None
--speed D	-s D	Learning speed	None
--iteration N	None	Counts of learning iteration (if data file is large)	None
--log N	None	Print info about learning each N iterations	None

If you launch without --output, then application will printing in stdout.

Install frontend

For example, i'm using apache
Frontend works without PHP and DB

$ cd /path/to/apache2 # Ubuntu: /var/www
$ ln -s /path/to/neural-network
$ chmod o+x neural-network
$ sudo -i
$ echo "\n127.0.0.1 yourdomain.loc" >> /etc/hosts
$ cd /etc/apache2/sites-available/
$ cp 000-default.conf nn-frontend.conf
$ nano nn-frontend.conf
### change domain and destination folder ###
$ a2ensite nn-frontend
$ systemctl restart apache2